How Orchard uses AI and machine learning to radically simplify homebuying and selling

There are few things more important to a homebuyer or seller than their home’s value. For buyers, this number impacts the amount they need for a down payment and, of course, the size of their future mortgage. For sellers, it represents the value of one of their largest financial assets and how much money they will (or won’t) walk away with in their pocket.

But determining a home’s value is a difficult task, requiring a delicate calculus of various home features, location factors, and market influences. It’s a job that is typically left to real estate professionals like realtors and appraisers, and one that online brokerages and iBuyers have been working tirelessly to automate.

While the prevalence of these online valuations is widespread, not all valuations are as precise as the next. Orchard’s home value estimator is 20% more accurate — so how did it get so, well, accurate? Meet the minds behind our patented automated valuation model (AVM).

But first, what’s an AVM? These software-based tools combine statistical modeling with existing databases of real estate data and transactions to calculate property values. It’s similar to a real estate agent’s comparative market analysis (CMA), which compares the values of similar properties at the same time, but AVMs streamline and automate the process.

The software is used to support a host of mortgage activities, like underwriting, home equity loans, refinancing, and loss mitigation. They’re also a key component of property tech companies like Orchard (as well as others like Zillow). While a licensed home appraiser is the only person who can deliver the final valuation of a home, AVMs are an essential part of the process, and are used in conjunction with the work of an in-house Pricing Team. As such, the AVM’s accuracy is paramount.

Orchard’s Data Science team has devoted years of research and development to honing their AVM. We spoke to Trevor Patch, Senior Manager of Data Science and Business Intelligence, to understand a little more about the team behind Orchard’s online estimator.

Patch, who recently celebrated his two-and-a-half-year anniversary with Orchard, says the Data team’s mission statement is simple: “Get the right data, to the right people, at the right time.” But the execution can be a little more complicated. 

“Orchard is three different companies in one — we’re a brokerage, a mortgage company, and a title company. So we have to provide a lot of different data to a lot of different, very smart stakeholders.”

That can mean sifting out high-intent leads from low-intent leads for the Sales team, or getting the Marketing team information on the performance of their latest campaign. But, perhaps, the most vital function is supporting Orchard’s AVM — a task easier said than done.

“We take a bunch of MLS data, tax data, and geographic data and track a home to about 70 different positive or negative adjacencies — things like cemeteries, power lines, airports, highways, rivers, and lakes, and parks.”

This is similar to what an appraiser or real estate agent might do in their CMA, but Orchard’s AVM doesn’t stop there. 

“We also do image classification work with listing photos. We classify the photos to rooms, then pull out features in those rooms. So, in the kitchen, does it have an island? What kind of countertops does it have? What kind of refrigerator? It’s kind of our special sauce.”

A sauce so special that its inventors — Mark Jayne, Gergely Svigruha, Raymond Liang, and Greg Aponte — were recently awarded a patent. Patch highlights how the patent is more than just a technical feat — it shows how seriously Orchard takes machine learning and AI. 

“A lot of companies talk about machine learning and AI applications, especially now that it’s the sexy thing to talk about, but very few companies have patents to show these things have always been core tenets of Orchard.”

And it’s a focus Patch has pursued ever since graduating. Before coming to Orchard, Patch built the Data Science and Software Engineering teams behind the Colorado Rockies. That’s right, he was the Moneyball guy.

While racking up W’s in pursuit of the World Series might seem like a far cry from radically simplifying the way that homes are bought and sold, there is one obvious carryover from Patch’s eight years at the Rockies: his focus on teamwork. He was adamant in highlighting the contributions of Raymond Liang and Winnie Ng on the Data Science team and Obie Odom, Melissa Lee, and Kyle Sullivan on the Business Intelligence team.

That’s another thing that Orchard can boast: playing for the name on the front of the jersey, not the back. If you’re interested in joining our team, get started by checking out our open roles today.